Ahmet Öztopal, Albero Mugnai, Daniele Casella, Marco ... - H-SAF

All what have been said about the two filtering cases are reflected in Figure 5.13 inthe form of spatial patches of rainfall occurrences and their quantities as shown in thecolor bars. It is obvious that the occurrence patterns are almost the same, but as for therainfall amounts are concerned, there are discrapencies. For instance, in the retrival case(Figure 5.13.a) although there are several sub-regions (peaks) with comparatively highrainfall amounts, radar case (Figure 5.13.b) has only a single peak.ab,,Figure 5.13. Patterns by using 3X3 filter on 8 February 2009 at 05:41 GMT (a) retrieval,(b ) radarFigure 5.14 is valid for 3 October 2008 at 05:44 GMT rainfall event both for retrivaland radar calculations. It is again obvious in Figure 5.15.a that there is a great scatter onthe basis of one-cell resolution and similarly overestimation still prevails, but comparedwith the case in Figure 5.13 the scatter are not very much concentrated on a certainportion on the Cartesian coordinate system. Comparison between the two cases indicatethat 3 October 2008 at 05:44 GMT event has less data points. On the other hand, coarserresolution (3x3 spatial filtering) yields better representation but still overestimation showsitself.Rain Rate Radar (mm/hr)3530252015105y = 0.7059x + 0.3093R 2 = 0.5026Rain Rate Radar (mm/hr)1512963y = 0.6038x + 0.2176R 2 = 0.8488a00 5 10 15 20 25 30 35Rain Rate Retrieval (mm/hr)b00 3 6 9 12 15Rain Rate Retrieval (mm/hr)Figure 5.14. Scatter diagrams between radar and retrieval on 3 October 2008 at 05:44GMT (a) original comparision, (b) after applying a 3X3 filter.21